// philosophy

Forensic skepticism by default

┌─────────────────────┐ │ AGENT GENERATES │ │ ▸ Questions │ │ ▸ Hypotheses │ │ ▸ Red flags │ │ ▸ Contradictions │ └─────────────────────┘

Agents ask, humans decide

No agent produces buy/sell signals in isolation. They surface patterns, anomalies, and questions that a human analyst would miss across 10,000 pages of filings.

if trust_score < 40: verdict = "AVOID" ← absolute if rpt_risk == "critical": verdict = "AVOID" ← no override # no benefit of the doubt # forensic skepticism is default

Forensic override rule

When forensic agents detect critical governance failures, the system forces an AVOID verdict. No philosophy lens, no growth narrative, no special situation can override fraud signals.

Model A ──┐ Model B ──┤──▸ Same Data Model C ──┤──▸ Different Lens Model D ──┘──▸ New Questions iterate(data, agents, models)

Iterate with different minds

The same OSINT data gets re-examined by different specialist models, different agents, different critical viewpoints. What Claude misses, Gemini catches. What a value investor ignores, a forensic accountant flags.

// agents

55 agents across 7 tiers

T1 Philosophy Lens 6
Growth Compounder Deep Value / Graham Bakshi Special Situations Contrarian / Mean Reversion Munger Latticework Technical Analysis
T2 Forensic & Quality 10
Earnings Quality Accounting Forensic Capital Allocation Promoter Behavior Management Track Record Board / Governance Mgmt Compensation Mgmt Commentary RAG Social Media Monitor Analyst Credibility
T3 Macro & Cycle 9
Interest Rate / Liquidity Credit Cycle Currency Exposure Commodity Cycle Bond / Debt Monitor Election / Political Regulatory Regime Retail Behavioral / Herd Retail Sentiment
T4 Event-Driven 8
Event Arbitrage Earnings Surprise Predictor IPO / Listing Analysis Filing Monitor News Aggregator Management Change Legal Case Tracker Pre-IPO Pipeline
T6 Portfolio-Level 6
Portfolio Construction Pessimist / Risk Scenario Engine Consensus / Synthesis Decision Journal Tax & Compliance
T7 Data Infrastructure 3
Data Quality / Reconciliation Insider Buying / Selling Non-Core Assets / Hidden Value
SP Sector Specialists 4
Insurance IT Services NBFC / Lending Real Estate / REITs
TR Unique Trackers 2
Zombie Company Detector Sector Deep-Dives
// signal_frequency

Two bands of intelligence

HIGH FREQUENCY

Real-time to weekly. News, social sentiment, price anomalies, management statements, insider trades, regulatory notices.

Triggers, not conclusions.

HF confirms LF→ Conviction ↑
HF contradicts LF→ Re-examine thesis
HF without LF base→ Noise risk
LF without HF trigger→ Early warning ★
LOW FREQUENCY

Quarterly to annual. Financial statements, cash flow, RPT disclosures, promoter pledging, capital allocation, board composition.

Builds conviction.

// agent_feed

Always running, always questioning

lattice agent-feed --live
09:31:04 PROMOTER MARUSHIKA TECH CRITICAL Longthrive Capital (anchor investor) incorporated 2 months before allocation. DIN network overlap with promoter.
09:31:07 FORENSIC MARUSHIKA TECH CRITICAL DGGI investigation u/s 74 (fraud/wilful misstatement) for fake invoice trading. Not prominently disclosed in DRHP risk factors.
09:31:12 RPT_MONITOR MARUSHIKA TECH CRITICAL ₹17.5Cr permanently parked with promoter entities. ₹19Cr circular loans with MD. Cash conversion ratio: 12%.
09:31:15 IPO_ANALYSIS MARUSHIKA TECH HIGH Zero mutual fund / insurance / pension participation in anchor allocation. 100% anchor = unknown micro-fund.
09:31:18 SYNTHESIS MARUSHIKA TECH AVOID Forensic override triggered. trust_score=12, rpt_risk=critical. 38 flags (15 critical, 15 high, 8 medium).
────────────────────────────────────────────────────────
09:32:01 VALUE TATA MOTORS MEDIUM P/E compressed to 7.2x vs sector median 14.8x. Graham net-net screen: passes on tangible book. But — check JLR debt covenants.
09:32:04 MACRO SECTOR:AUTO INFO RBI policy rate cut 25bps. Auto loan disbursement up 18% YoY. But rural demand index flat — urban-rural divergence widening.
09:32:09 EARNINGS_Q YES BANK HIGH Cash flow from operations negative for 5th consecutive quarter despite reported profit growth. Accrual vs cash divergence: 340%.
09:32:14 GOVERNANCE CIAN AGRO HIGH Promoter holds political office while controlling 72% float. Independent directors appointed post-election. Board capture risk.
█
// whitespace

What doesn't exist yet

55 > 10

Agent scale

Largest documented multi-agent financial system uses ~10 agents. ZenoLattice: 55.

■■■■■

Forensic + Investment

No system combines forensic accounting, governance scoring, and investment analysis in a single multi-agent LLM framework.

SEBI MCA BSE NSE

India regulatory intel

31 agents with explicit India-specific edge. SEBI filings, MCA data, BSE/NSE structures, RBI circulars, Indian promoter patterns.

RPT → DIN → ★

RPT + Promoter AI

Related-party transaction monitoring and promoter behavior analysis via AI — virtually unaddressed globally.

↻ iterate()

Multi-model iteration

Same data, different models, different agents, different epochs. New model drops? Re-run everything. New agent? Plug in and iterate.

// early_access

Get forensic intelligence
before the market does

We're onboarding a small group of institutional investors, proprietary desks, and financial journalists for early access to ZenoLattice signals.

Mutual Fund Houses Proprietary Investors PMS / AIF Managers Financial Journalists Equity Research Analysts
We'll share sample forensic reports and signal feeds. No spam, no selling your data.